基于词向量和深度学习模型的医疗数据分析方法研究  

Research on Medical Data Analysis Method Based on Word Vector and Deep Learning Model

作　　者：金玮[1] 左嵩[1] 许健[1] 黄于飞 巩清源[1] 潘伟华[1] JIN Wei;ZUO Song;XU Jian;HUANG Yufei;GONG Qingyuan;PAN Weihua(Department of Information Management,Xinhua Hospital Affiliated to Shanghai Jiaotong University School of Medicine,Shanghai 200025,China)

机构地区：[1]上海交通大学医学院附属新华医院信息管理部,上海200025

出　　处：《微型电脑应用》2021年第5期23-26,共4页Microcomputer Applications

基　　金：上海市科委科研计划项目(16411962601)。

摘　　要：在自然语言处理领域中,电子医疗数据越来越多地与文本处理技术结合使用,以辅助医生进行诊断。但电子医学文本数据通常存在大量冗余、语义缺失、歧义等问题,导致基于特征提取的传统分类器不能充分发挥作用。针对上述问题,结合词向量法与深度学习理论,将医学文本中的数据和词表示为一个定长的矩阵,并利用改进LSTM模型与改进Yoon模型进行结合,整合成一个综合学习模型。最终,在实际医疗数据分类实验中验证了综合学习模型的有效性。实验与数据分析结果表明,对比改进后的LSTM模型与Yoon模型,综合学习模型在医学电子数据文本分类的准确率及精确度分别得到了较为明显的提高。In the field of natural language processing,electronic medical data are increasingly used by combining with the text processing technology to assist doctors in diagnosis.However,electronic medical text data often have a large amount of redundancy,lack of semantics,ambiguity,etc.,which causes traditional classifiers based on feature extraction cannot fully implement function.In response to the above problems,the article combines word vector method with deep learning theory,expresses the data and words in medical text as a fixed-length matrix,and combines the improved LSTM model with the improved Yoon model to form a comprehensive learning model.Finally,the effectiveness of the comprehensive learning model is verified by the actual medical data classification experiment.The results of experiments and data analysis show that comparing the improved LSTM model and the Yoon model,the accuracy and precision of the comprehensive learning model in medical electronic data text classification have been significantly improved.

关 键 词：词向量 深度学习 电子医学文本分析 LSTM模型 Yoon模型 

分 类 号：TP391[自动化与计算机技术—计算机应用技术]